World-Wide Web ( WWW ) becomes extremely popular in recent years. However, there still exists many fundamental problems not yet solved in a satisfactory way, such as the information search, distribution, caching, filtering problems, etc.. Users still often suffers from long retrieval latency, and the lack of search function which can conduct localized search or hierarchical search. In this paper, we propose a framework for solving the above mentioned fundamental problems by introducing a new model of WWW proxy servers. In our model, a proxy is not just an add-in to existed Web servers to provide cache and security functions, but works as a middle layer between browsers and Web servers, which can be plugged in a variety of functions for WWW applications, such as search engine, resource index, information filtering, robot, profile, intelligent agents, etc.. It's a framework for implementing hierarchical information servers and it's the right place to plug in various intelligent agents. To illustrate the usefulness of our model, we have implemented two fundamental functions into this middle layer : the object lookup scheme (collaborative caching) and the searchable proxy. We call this extensible proxy server "Virtual Proxy" . Since search and retrieval is the most fundamental problems in World Wide Web, we will present this paper in a way to show how these two problems can be solved in a better way using the Virtual Proxy servers.